A strong building is not defined by what you see above the ground. It's defined by what was done below it. #CivilEngineering #ConstructionLife #EngineeringLessons #ProjectManagement #BuildBetter #Uganda

Every new capability introduces new dependencies. Managing those dependencies is becoming a strategic requirement. #EngineeringLessons

One thing construction has taught me is that progress is rarely visible every day. Not all progress is visible. Not all growth is obvious. Just Keep building. #EngineeringLessons #ConstructionLife #CivilEngineering #GrowthMindset

When 1 isn’t equal to 1. Opening our Engineering Lessons series with Mars Climate Orbiter, lost after a mismatch between imperial and metric units. A small inconsistency. A major consequence. Credit: NASA #EngineeringLessons #SpaceTech

Machine Learning IoT is not plug-and-play. I learned this the hard way. Most posts make Machine Learning and IoT look simple. Connect a sensor. Train a model. Deploy it. Done. That version exists only in slides. When I started building real IoT systems using ESP32 and STM32 with Machine Learning, reality pushed back hard. I learned one uncomfortable truth early: Models do not fail first. Systems do. Below are the real failures, fixes, and lessons that shaped how I now engineer ML-powered IoT systems. Failure 1: A perfect model that collapsed in the field In controlled conditions, the model performed well. Once deployed: - Sensor drift appeared - Environmental noise increased - Installation conditions changed The algorithm was not wrong. The assumptions were. What fixed it: - Collecting real field data instead of clean lab data - Normalizing inputs directly on the device - Rejecting invalid and unstable sensor readings Lesson: Data quality matters more than model complexity. Failure 2: A powerful model on constrained hardware A model that runs smoothly on a PC does not automatically belong on a microcontroller. - On ESP32 and STM32 hardware: - Memory became the bottleneck - Inference latency increased - Power consumption mattered What fixed it: - Choosing simpler models over complex ones - Using quantized models instead of floating point - Running ML only where it created real value Lesson: In IoT, the best model is the one that fits the hardware. Failure 3: Internet dependency destroyed reliability Early designs assumed constant connectivity. Reality delivered network drops, latency spikes, and unstable links. The system became fragile. What fixed it: - Edge-first system design - Local fallback logic Using ML for prediction, not for safety-critical control Lesson: Machine Learning should increase reliability, not depend on ideal conditions. Failure 4: High accuracy with low real-world impact I once celebrated a model with strong accuracy metrics. Stakeholders did not. Accuracy did not reduce: - False alarms - Maintenance cost - System downtime What fixed it: - Defining success at the system level - Optimizing for stability and reliability - Letting engineering logic lead and ML support Lesson: Accuracy is a metric. Impact is the goal. What finally worked Simple models instead of complex ones Real-world data instead of synthetic perfection A balance between edge and cloud Hardware-aware ML design Combining ML with rule-based control Final thought Machine Learning in IoT is not about smarter models. It is about building systems that survive reality. That mindset shift, from writing code to engineering systems, changed how I build everything. 🔥 Top 6 ESP32 eBooks ebokify.com/top-6-esp32-eboo… #MachineLearning #IoT #EdgeAI #EmbeddedSystems #ESP32 #STM32 #TinyML #AIEngineering #EngineeringMindset #SystemsEngineering #RealWorldAI #HardwareMeetsSoftware #EngineeringLessons #LearningByBuilding

Yesterday’s “quick hack” is today’s technical debt. Before you refactor anything, write down the constraint that forced the hack in the first place. Understanding beats blaming. What old hack are you turning into a lesson this week? #EngineeringLessons #MorningThoughts

#BuildInPublic #SoftwareEngineering #FinTech #TeamWork #EngineeringLessons #DevOps

Engineering Dilemmas: Navigating the Balance Between Innovation and Safety As an engineer, I've learned the importance of anticipating potential problems and taking preventive measures in every project. Risk investigation and mitigation are key. #EngineeringLessons #Titan

Can’t wait till this is me 🥺 I love engineering so much 😩 shoutout to you if you follow me on snap and watch the #engineeringlessons I give lol ☺️

Build Design challenge @MIT InvenTeams #PD loving the #EngineeringLessons #ConnectingEducators

TECET hosts @BostonGlobe Reporter @neilswidey TRAPPED UNDER THE SEA on 5/29, #engineeringlessons bit.ly/NeilSwidey_TECET #Registertoday